It is desired to realize a solid-state imaging device having a wide dynamic range in which high-luminance information does not collapse even in backlight imaging against a light source such as headlight of a vehicle, light illuminating a stadium, or sunlight, and a detailed part can be imaged without excessively darkening a low-luminance part of a subject image.
In such circumstances, techniques for enlarging the dynamic range of a solid-state imaging device such as a CCD are disclosed in Japanese Patent No. 2125710, JP-A-03-117281, JP-A-09-205589, JP-A-2004-320119, and the like.
In Japanese Patent No. 2125710, an example is represented in which a plurality (at least two) of areas (cells) having different sensitivity characteristics that correspond to a knee or more are disposed within one pixel such as a CCD, and the dynamic range of the CCD is increased by implementing so-called knee characteristics in which the input/output characteristics change in a stepped manner.
The knee characteristics refer to a characteristic curve in which a curve represented as the relation of an output current with respect to the exposure amount is lower in a high-input region than in a low-input region and are frequently referred to as a high-luminance signal compressing technique.
As methods of changing the sensitivity of a photosensitive area (cell), for example, a method in which the aperture ratio of the device is changed, a method in which an optical filter (ND filter) is disposed, a method in which the impurity density is changed, and the like are disclosed.
In Japanese Patent No. 2125710, although the technique is described to be applicable to an XY address-type imaging device other than the CCD, there is no detailed description.
In JP-A-03-117281, an example is disclosed in which, signal electric charges of cells disposed within one pixel are added so as to be set as the signal electric charge of the cell with pixels adjacent to each other or cells having mutually different photosensitive characteristics among photosensitive pixel cells of a CCD configured as one set, and a high dynamic range is acquired for which information does not collapse even under high light of a bulb or the like.
In such a case, as a unit that changes the photosensitivity, for example, cells having mutually different pixel areas are configured as a set.
In JP-A-09-205589, in the same way, one pixel of the photosensitive pixel cell of a CCD is divided into two areas having mutually different photosensitivity levels, and the signal electric charges of the areas in which the photosensitivity of the same pixel differs from each other are mixed and are vertically transmitted. Then, according to this technique, signal charges having mutually different photosensitivity levels are distributed to two horizontal transmission gates by a distribution gate, and a high-sensitivity side signal is clipped by an external signal processing circuit and then is added to a low-sensitivity side signal so as to form a video signal.
In such a case, the characteristic graph of the video signal output with respect to the intensity of incident light has the shape of a polygonal line, the gradient of the high sensitivity side (low illumination side) is sharp, and the gradient of the low sensitivity side (high illumination side) is gentle.
In JP-A-2004-320119, a method of solving a problem is disclosed in which, in an imaging device having a high-sensitivity cell and a low-sensitivity cell, the amount of data (raw data) of a raw image is large due to the data of both cells.
More specifically, it is automatically determined whether it is necessary to record image information of a high illumination unit by analyzing captured image information. In a case where “Yes” is determined, raw image data of the high illumination unit is recorded together with the information of a low illumination unit. On the other hand, in a case where “No” is determined, only the raw image data of the low illumination unit is recorded without recording the information of the high illumination unit.
A main photosensitive pixel cell (high sensitivity due to large area; a center portion of a micro-lens is mainly used) and a sub-photosensitive pixel cell (low sensitivity due to a small area; it is disposed on the edge side of the micro-lens) are combined so as to form one pixel.
In JP-A-2005-278135, a CMOS image sensor is disclosed in which a column-parallel ADC is configured by a comparator and an up/down counter. This CMOS image sensor can perform an addition calculation of pixel digital values over a plurality of rows without adding additional circuits such as an adder and a line memory device.
However, in the case of the above-described addition of division pixels, compared to a pixel having an area corresponding to a total area of target pixels, an ineffective area (dead space) that does not directly contribute to exposure is generated due to signal processing in a case where the pixel is divided.
Accordingly, since the area of an individual divided cell is smaller than that of a case where the cell is simply divided into four division cells, the number of saturated electrons is smaller than that of the former case, and accordingly, a shot noise relatively increases, whereby the S/N of the individual division pixel is degraded.
Whenever addition is performed, the shot noise is also added, and accordingly, the S/N of the result of the addition of division cells is also degraded.
In addition, the addition process of pixel signals is addition of analog signals, and the sensitivity differs for each pixel, and accordingly, there are problems that the saturated value is uneven, the break point position varies, and the like.
Furthermore, in the case of digital addition, it is necessary to include a memory in addition to the sensor.
In other words, in an existing addition method in which one pixel cell is divided into two or more pixel cells of which the sensitivity levels or accumulation times are different from each other, and the photosensitivity is measured as the amount Qs of saturated electric charge, there are variations in the amount Qs of saturated electric charge for each pixel. Accordingly, for the same light intensity, the addition result varies for each pixel.
In other words, in a sensitivity graph (polygonal graph) in which the intensity of incident light is set as the horizontal axis, and the amount Qs of saturated electric charge is set as the vertical axis, at a divided pixel cell adding point (horizontal axis), the break point position (vertical axis) varies.
Thus, in JP-A-2010-28423, a method of realizing a wide dynamic range by changing the accumulation times of four pixels by applying a technique for regarding four pixels as one pixel is proposed. In this technique, four signals are added together.
According to this technique, the addition of division pixels in which there is no variation in the numbers of output electrons of pixels with respect to the intensity of incident light can be realized, the sensitivity is increased when the intensity of incident light is low, and the sensitivity is decreased when the intensity of incident light is high, whereby a wide dynamic range for which it is difficult for the output to be saturated can be acquired.